Is it actually a draw?

[syzygy]

OK. But if the shortest win here needs say 49 moves, then this line is valueless in the position as originally posted. Since I have no idea what M1000 from H4 specifically means, I'm not sure that seeing it guarantees a win in the position as originally posted.

What does SF+Syzygy give as evaluation here?

[D]8/5k2/5r2/1p3q2/5B2/7R/6QK/8 w - -

My engine says win in 15 moves.

Code: Select all

+Win15   1.Qa2+ Kg7 2.Rh4 b4 3.Qa7+ Kg8 4.Qb8+ Rf8 5.Qxb4 Rf6 6.Qb8+ Kf7 7.Qc7+ Ke6 8.Qc4+ Kd7 9.Kg3 Rg6+ 10.Kf2 Rc6 11.Qd4+ Ke8 12.Rh8+ Kf7 13.Qa7+ Ke6 14.Re8+ Kd5 15.Re5+ Kc4 16.Rxf5

or another run:

Code: Select all

+Win15   1.Rh4 b4 2.Qb7+ Ke6 3.Qa6+ Kf7 4.Qc4+ Qe6 5.Qc7+ Kg8 6.Qh7+ Kf8 7.Qh8+ Kf7 8.Rh7+ Kg6 9.Rg7+ Kf5 10.Qh4 Qc6 11.Rg5+ Ke6 12.Qh7 Qc4 13.Re5+ Kd6 14.Re4+ Kc5 15.Rxc4+ Kb5 16.Rxb4+

Some might prefer to say +Win16 here .

Note that the losing side will refuse to (re)capture into a losing 6-piece position if it can (i.e. does not see that it gets mated otherwise), so as soon as you get large scores it means white knows it can gain material and black won't dare to take back. The winning lines you see here have two captures by white.

[Houdini]

Code: Select all

+Win15   1.Qa2+ Kg7 2.Rh4 b4 3.Qa7+ Kg8 4.Qb8+ Rf8 5.Qxb4 Rf6 6.Qb8+ Kf7 7.Qc7+ Ke6 8.Qc4+ Kd7 9.Kg3 Rg6+ 10.Kf2 Rc6 11.Qd4+ Ke8 12.Rh8+ Kf7 13.Qa7+ Ke6 14.Re8+ Kd5 15.Re5+ Kc4 16.Rxf5

The Qa2+ line is most peculiar, compared to the initial diagram of this thread the only change is that the white Rook has moved from f3 to h3 in the space of 10 moves.

This is a truly remarkable ending. Black is in some kind of zugzwang and cannot leave the f-file with his pieces. Beautiful!

[Houdini]

P.S. for Milos: It is the deep zugzwang nature of this position that makes it difficult for engines to recover the PV when some of the hash information is lost.

[Milos]

P.S. for Milos: It is the deep zugzwang nature of this position that makes it difficult for engines to recover the PV when some of the hash information is lost.

I did retrograde analysis starting from positions where white captures a piece in PV (i.e. score 168 or 318) with sufficient depth. Once this info is in hash it becomes much easier for engine to prune the lines, so going backwards it manages to remove all the refutations. I put hash to 16GB but it is still not enough, however, managed to trace back to move 50.
We don't have 7-men (black B pawn is irrelevant anyway), but it seams at that point black position is lost. Simply starting from that position 7-men would show the win for white. It is done in a way that white pushes black into deep zugzwang and eventually captures a piece (or takes Q for R).
It seams black really lost the game on move 36 when it didn't exchange queens. However, even for H4 it takes few hours to see that.
It is truly a remarkable endgame.

[jhellis3]

You may also want to check why Fire did not expect 16. Qxg2.

[peter]

Play the PV moves to get more insight.
From your initial diagram: 1...Kg7 2.Qa7+ Kg8 3.Rg3+ Rg6 4.Qb8+ Kf7 5.Qc7+ Kg8 6.Rh3 Rf6 7.Qb8+ Kf7 8.Qb7+ Kg8 9.Qg2+ Kf7
[D]8/5k2/5r2/1p3q2/5B2/7R/6QK/8 w - -
You will find a "M1000" score which indicates a Syzygy win.
Both Bg3 and Rh4 appear to be winning, quite a remarkable ending.

As much as that you can say for sure, Robert!

After your variant and 10.Bg3 with 6men Nalimov's:


8/5k2/5r2/1p3q2/8/6BR/6QK/8 b - -

Engine: Houdini 4 Pro x64L (8192 MB)
von Robert Houdart

27 73:33 +M36 10...De6 11.Db7+ Kg8 12.Dh7+ Kf8
13.Dh8+ Ke7 14.Th7+ Tf7 15.Lh4+ Kd7
16.Dd8+ Kc6 17.Da8+ Kc5 18.Da3+ Kd4
19.Da1+ Kd3 20.Dd1+ Kc4 21.Dc2+ Kb4
22.Le1+ Dxe1 23.Txf7 (47.703.544.818) 10808
TB:32.244.288

27 73:33 +M32 10...b4 11.Th8 Kg7 12.Th4 Kf7
13.Db7+ Kg8 14.Da8+ Tf8 15.Dg2 Dg6
16.De2 Df5 17.Tg4+ Kf7 18.Dc4+ Kf6
19.Lh4+ Ke5 20.Db5+ Ke6 21.Dc6+ Ke5
22.Lg3+ Df4 23.Lxf4+ Txf4 (47.703.544.818) 10808
TB:32.244.288